The outcome

Answer from authorized sources with evidence and abstention without leaking files or granting an extension excessive access.

LobeHub is a Open-source AI workspace centered on LobeChat, with multi-provider chat, agents, knowledge bases, files, plugins, MCP, and self-hosting. Using multiple AI providers in one workspace, building reusable assistants with files and knowledge, and operating a private multi-user LobeChat service with controlled extensions. This guide narrows that broad capability into one repeatable outcome, with checkpoints that keep the source material and your judgment in the loop.

Before you begin

Set the boundary before the tool starts.

Choose one real task, identify who will use the result, and decide what evidence or test will make the result acceptable. Gather only the source material needed for that task. If the work contains confidential, personal, regulated, or client-owned information, confirm that the platform and account are approved before sharing it.

Troiana principle

AI should make the work easier to inspect. If the workflow removes the source, the owner, or the review step, redesign the workflow.

Step by step

A workflow you can repeat.

  1. 01

    Define audience, source owners, document classes, access rules, freshness, answer and citation requirements, model and embedding providers, tools, retention, and approval boundaries.

  2. 02

    Create an isolated knowledge collection, ingest minimal approved files, remove secrets and unnecessary personal data, and record owner, version, rights, tenant, and expiry metadata.

  3. 03

    Configure a narrow assistant with evidence-first instructions and abstention, bind knowledge to the correct user or tenant, and add only read-only plugins or MCP tools from reviewed sources.

  4. 04

    Test retrieval relevance, missing and conflicting evidence, stale sources, cross-tenant IDs, document prompt injection, malicious tool output, oversized files, provider failure, and unauthorized actions.

  5. 05

    Review citations and traces, reindex approved updates, audit tool and knowledge permissions, monitor provider cost, and support source removal, user export, deletion, and immediate extension disable.

Working standard

What good use looks like.

  • Bind knowledge to authenticated tenants.
  • Require evidence or abstention.
  • Vet every plugin and MCP server.

Self-hosting the interface does not make remote model, embedding, search, storage, sync, or plugin traffic local. A database deployment adds authentication, database, vector, object-storage, migration, backup, and patch responsibilities. Never expose an unauthenticated instance or provider key; isolate tenants and knowledge, vet plugins and MCP servers, restrict registration and networks, redact observability, and test restores before upgrades.

Official references

Check the current product documentation.

Features, plan limits, availability, and data controls change. These official pages are the starting points used for this collection.